Impact of Calibration Rainfall Limits on the Accuracy of Curve-Number Models

Methods for estimating curve numbers from field measurements are investigated. It is shown that for the same calibration data set the curve number (CN) estimated using the median method will generally be greater than the CN estimated using the least-squares method. Errors in estimating runoff using curve numbers derived using different calibration methods will generally depend on both the range of rainfall-runoff measurements used in calibration and the rainfall amount for which runoff is to be predicted. It is shown that substantial errors can occur when predicting runoff from catchments with calibrated asymptotic curve numbers (CN∞) of around 40. For catchments with calibrated CN∞ of around 60, the potential for runoff error diminishes, but concerns remain particularly for rainfall amounts less than around 5 cm. For catchments with calibrated CN∞ of around 80, there is minimal potential for error regardless of the rainfall amount.